The Electric Multiple Unit (EMU) is a sophisticated asset renowned for its high reliability. implementation of advanced health management technology can enhance the operational reliability and safety EMUs. Preventive replacement interval planning, key aspect EMU lifetime management, requires significant investment in funds, manpower, resources. However, it has potential to substantially reduce maintenance expenses improve profitability operations. This paper introduces method estimating...

Solid state drives (SSDs) have been widely deployed in high performance data center environments, where multiple tenants usually share the same hardware. However, traditional SSDs distribute users' incoming uniformly across all SSD channels, which leads to numerous access conflicts. Meanwhile, that statically allocate one or several channels tenant sacrifice device parallelism and capacity. When are shared by with different patterns, inappropriate channel allocation results degradation. In...

Abstract We introduce a new method based on nonnegative matrix factorization, Neural NMF, for detecting latent hierarchical structure in data. Datasets with arise wide variety of fields, such as document classification, image processing, and bioinformatics. NMF recursively applies layers to discover overarching topics encompassing the lower-level features. derive backpropagation optimization scheme that allows us frame neural network. test synthetic dataset, 20 Newsgroups MyLymeData symptoms...

Emerging applications like machine learning in embedded devices (e.g., satellite and vehicles) require huge storage space, which recently stimulates the widespread deployment of large-capacity flash memory IoT devices. However, existing file systems fall short managing efficiently for excessive consumption poor booting performance. In this paper, we propose a novel system, LOFFS, to tackle above issues manage NAND on resource-limited We redesign space management mechanisms construct hybrid...

Emerging Zoned Namespace (ZNS) is a new-style Solid State Drive (SSD) that manages data in zoned manner, which can achieve higher performance by strictly obeying the sequential write mode each zone and alleviating redundant overhead of garbage collections. Unfortunately, flash memory usually has serious problem with limited program/erase cycles. Meanwhile, inappropriate placement strategy storage systems lead to imbalanced wear among zones, severely reducing lifespan ZNS SSDs. In this paper,...

Solid-state drives (SSDs) have been widely deployed in high-performance data center environments, where multiple tenants usually share the same hardware. However, traditional SSDs distribute users’ incoming uniformly across all SSD channels, which leads to numerous access conflicts. Meanwhile, that blindly allocate one or several channels tenant sacrifice device parallelism and capacity. When are shared by with different patterns, inappropriate channel allocation results performance...

Emerging applications like machine learning in embedded devices (e.g., satellites and vehicles) require huge storage space, which recently stimulates the widespread deployment of large-scale flash memory IoT devices. However, existing file systems fall short managing large-capacity efficiently for two reasons. First, prior arts store data structures either or main memory, severely magnifies scarcity computing resources. Moreover, fine-grained metadata management induces significant energy...

Index structures can significantly accelerate the data retrieval operations in intensive systems, such as databases. Tree structures, B <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">+</sup> -tree alike, are commonly employed index structures; however, we found that tree structure may not be appropriate for Non-Volatile Memory (NVM) terms of requirements high-performance and high-endurance. This paper studies what is best NVM-based systems how...

The mismatch between logical and physical I/O granularity inhibits the deployment of embedded file systems. Most existing systems manage space with a small unit, which is no longer case flash operation granularity. Manually enlarging requires enormous transplanting efforts. Moreover, large pages signify write amplification problem, turns to severe consumption performance collapse. This article designs novel storage middleware, NV-middle, for legacy-embedded large-capacity memories....

Existing persistent memory file systems can significantly improve the performance by utilizing advantages of emerging Persistent Memories (PMs). Especially, they employ superpages (e.g., 2MB a page) PMs to alleviate overhead locating data and reduce TLB misses. Unfortunately, superpage also induces two critical problems. First, consistency using causes severe write amplification during overwrite data. Second, existing management may lead large waste PM space. In this paper, we propose...

With the development of embedded devices, elevator group systems that manage elevators can be designed in an intelligent way. In design systems, one most important problems is to determine "range assignment" for each elevator, which indicates floors will serve. reality, traffic loads a building are different terms time periods, called fluctuant loads, makes above problem much more challenging. The objective this paper optimal range assignment maximize number passengers served certain amount...

Softmax policy gradient is a popular algorithm for optimization in single-agent reinforcement learning, particularly since projection not needed each update. However, multi-agent systems, the lack of central coordination introduces significant additional difficulties convergence analysis. Even stochastic game with identical interest, there can be multiple Nash Equilibria (NEs), which disables proof techniques that rely on existence unique global optimum. Moreover, softmax parameterization...

Distributed optimal control is known to be challenging and can become intractable even for linear-quadratic regulator problems. In this work, we study a special class of such problems where distributed state feedback controllers give near-optimal performance. More specifically, consider networked with decoupled costs spatially exponentially decaying dynamics. We aim exploit the structure in problem design scalable reinforcement learning algorithm controller. Recent work has shown that...

Existing persistent memory file systems usually store inodes in fixed locations, which ignores the external and internal imbalanced wears of on (PM). Therefore, PM for storing can be easily damaged. solutions achieve low accuracy wear-leveling with high-overhead data migrations. In this paper, we propose a Lightweight Multi-grained Wear-leveling Mechanism, called LMWM, to solve these problems. We implement proposed LMWM Linux kernel based NOVA, typical system. Compared MARCH, state-of-theart...

Conventional journaling file systems, such as Ext4, guarantee data consistency by writing in-memory dirty inodes to block devices twice. The write back of may contain up 80% clean inode that is unnecessary be written back, which caused severe amplification problem and largely reduce performance since the size an several times less than a basic unit for updating device. Emerging persistent memories (PMs), phase change memory, provide possibility storing offset in memory persistently. In this...

Block-based file systems, such as Btrfs, utilize the copy-on-write (CoW) mechanism to guarantee data consistency on solid-state drives (SSDs). Open-channel SSD provides opportunities for in-depth optimization of block-based systems. However, existing systems fail co-design two-layer semantics and cannot take full advantage open-channel characteristics. Specifically, synchronizing an overwrite in Btrfs will all pages update path induce severe write amplification. In this paper, we propose a...

We have witnessed the tremendous success of deep neural networks. However, this comes with considerable memory and computational costs which make it difficult to deploy these networks directly on resource-constrained embedded systems. To address problem, we propose TaijiNet, a separable binary network, reduce storage overhead while maintaining comparable accuracy. Furthermore, also introduce strategy called partial binarized convolution binarizes only unimportant kernels efficiently balance...

Emerging Persistent Memories (PMs) are promised to revolutionize the storage systems by providing fast, persistent data access on memory bus. Therefore, file developed achieve high performance exploiting advanced features of PMs. Unfortunately, PMs have problem limited write endurance. Furthermore, existing space management strategies usually ignore this problem, which can cause that operations concentrate a few cells PM. Then, unbalanced writes damage underlying quickly, seriously damages...